In short
- Hierarchy helped compress information and coordinate decisions across people.
- AI adds actors that can move work across systems at machine speed.
- Those actors need shared context, company intent, available capacity and decision authority.
“AI-native” is often described through speed. Faster code, faster drafts, fewer people per task.
It also changes how many actors can make decisions inside a company at once.
A person reads a record.
An agent updates another.
A workflow moves the work forward.
Each can affect what the others should do next.
The human router
Hierarchy does many things, control among them. The one that matters here is coordination.
Five people in a room can share context directly. As the organization grows, the possible relationships multiply. Five people have 10 pairwise relationships. Fifty have 1,225. Five hundred have 124,750.
Only some of those relationships require direct communication. The calculation shows how quickly the possible connections grow.
Managers help reduce that coordination burden in two directions.
- Upward, they filter. A hundred observations become the ten their manager needs to see.
- Downward, they focus. Which priority matters? Which work should wait? Who can commit the team?
That compression preserves attention. It can also lose context.
The coordination trade-off
A signal climbs the chain. Each person summarizes it for the next. By the time it reaches someone with authority, a detail that changes the decision may have disappeared.
Removing a layer shortens that route. It also leaves someone else carrying the coordination work.
The engineer has to explain the release risk to Sales. Sales has to reconcile the customer promise with Delivery. Finance needs to understand which commitment changes the forecast.
Information can be available while the operating picture remains fragmented.
A team can remove a reporting layer and still need to reconcile the same customer promise, delivery constraint and financial consequence.
The reporting line changes.
The dependencies remain.
Someone needs a way to see them together.
The design problem remains: how does an organization reduce information loss while keeping the coordination work manageable?
More actors, more coordination
In McKinsey’s August 2026 survey, 40% of respondents at companies with more than $1 billion in annual revenue reported scaling agents in one or more functions. The figure describes that cohort.
Microsoft’s May 2026 Work Trend Index surveyed 20,000 knowledge workers using AI across 10 markets. Only 26% said their leadership was clearly and consistently aligned on AI.
More actors can move more work. Their actions also create dependencies for other actors.
A sales agent books a deal. A delivery agent schedules the kickoff. A success agent plans the onboarding.
A manager can catch that conflict. As actions multiply, catching every conflict through manual review becomes harder.
Deloitte’s June 2026 operating-model research argues that scaling AI requires coordination across leadership, work, capital and risk, with clear decision rights.
Those choices need to reach the people and agents carrying the work. Each needs to know which decisions it may make, which require review and where to escalate a conflict.
Routing is not enough
A support lead sees ticket volume spike on a new feature. The system surfaces it to the product manager before the next roadmap is locked.
The support lead can stay in the ticket queue.
The product manager can stay in the planning tool.
The relevant signal crosses the boundary.
Knowing tickets spiked still leaves a decision. Should the release pause? Should support coverage increase? Which commitment can the company change?
Three things shape the response.
- Intent. The targets, policies and priorities the company has chosen.
- Capacity. The people, agents and other resources available to carry the work the decision creates.
- Authority. Who can recommend, who can decide and who can commit the organization.
A manager often carries those together: the priority, the stretched team and the authority to say yes.
A routed signal needs that context to support the next decision.
A shared operating model
The front line sees what changed. A shared model connects the change to affected work, available capacity and the authority to respond.
The product manager can inspect the support pressure behind a proposed delay. The support manager can see which release commitment constrains the response. The person deciding has a common basis to work from.
Shared memory preserves the evidence. Intent, capacity and authority establish how the company can use it.
Some facts will be missing.
Some priorities will conflict.
Some decisions will still require a conversation.
The model should make those conditions visible. It gives managers a common picture to challenge and decide from, reducing the work of rebuilding that picture in every meeting.
What people are for
The system can connect a delayed launch, a full delivery team and a decision that belongs to a VP.
Someone still has to judge the customer relationship, negotiate competing commitments and decide which consequence the company is willing to accept.
People set intent.
They approve the policies.
They hold the relationships.
They handle exceptions the rules leave unresolved.
They remain accountable for the commitments they make.
Recording the decision and what followed gives the next review a basis. It also creates a question for the people who made it: did the response work, and what should change?
What this means for your team
Look for the places where coordination depends on one person assembling the picture:
- A recurring meeting mainly reconciles three versions of a spreadsheet.
- An operator knows which accounts need attention, and the review stalls when they are away.
- Teams add agents without a shared view of the work or capacity those agents affect.
- A decision gets reopened because nobody can establish what was known when it was made.
Each is a place to make the operating model more explicit. Start with the decision, the evidence it needs and the person who has authority over it.
The decision runtime
This is what OpX is built to be: the decision runtime for customer operations, starting in Customer Success.
It connects the work, the capacity to deliver it and the authority to decide. Its design gives people a shared basis for review. An external agent needs a defined scope and authority over any work it may carry out.
As more actors join the work, the company still needs to know which commitments it can support and who may make them.
Sources
- The state of AI in 2026: On the road to ROI, McKinsey, August 25, 2026.
- Agents, human agency, and the opportunity for every organization, Microsoft 2026 Work Trend Index, May 5, 2026.
- Rewiring the enterprise operating model for AI scale, Deloitte Insights, June 29, 2026.
